{"date":"2026-11-08","title":"What happened in AI — Nov 8, 2026","generated_at":"2026-11-08T22:15:23Z","intro":["Builders are shrinking their agent stacks: Lovable deleted most of its agent code and one engineer dropped all agent skills, citing Anthropic's 80%+ Claude Code prompt cut with no measurable eval loss.","The infrastructure gap is memory: every coding tool stores it differently, and several projects propose a shared format or an inspectable store."],"highlights":["Lovable deleted almost all its agent code after 100k PRs.","Anthropic reportedly cut over 80% of Claude Code's system prompt with no measurable eval loss.","Agent memory lacks a shared format; the OpenTelemetry analogy is gaining ground.","ARTEX-driven campaigns stole 190k gaming credentials for under $400 in model costs.","Uncensored models in agent loops: most fail about half of multi-step tasks."],"article_count":17,"categories":[{"name":"Agent harnesses: less scaffolding, more deletion","slug":"agent-harnesses-less-scaffolding","summary":"Builders report pruning agent code and skills as models improve, and the lighter harnesses they keep are small and inspectable.","articles":[{"title":"Lovable: we deleted almost all our agent code","summary":"Lovable's lessons from 100k PRs: choices that once helped can hold you back, so the team keeps rebuilding how the agent works.","source":"hackernews_ai","url":"https://lovable.dev/blog/how-we-built-lovable-and-why-we-keep-rebuilding-it","published":"Sun, 08 Nov 2026 16:09:37 +0000"},{"title":"Author deletes all their agent skills","summary":"Cites Anthropic cutting over 80% of Claude Code's system prompt with no measurable loss on its own evals, and argues skills can over-constrain the agent.","source":"hackernews_ai","url":"https://philippdubach.com/posts/i-just-deleted-all-my-agent-skills/","published":"Sun, 08 Nov 2026 15:04:51 +0000"},{"title":"psi: a 6.4 MB coding agent in ANSI C and Lua","summary":"MIT-licensed core in C with the agent loop, tools and providers in Lua; works with Anthropic, Codex, OpenRouter or Ollama.","source":"hackernews_ai","url":"https://psicode.ai/","published":"Sun, 08 Nov 2026 17:48:19 +0000"},{"title":"My Local Coding Agent","summary":"A 25-year programmer describes a local agent setup and says LLMs in agent harnesses became capable of full features in January 2026.","source":"hackernews_ai","url":"https://hdembinski.github.io/posts/coding_agent.html","published":"Sun, 08 Nov 2026 19:12:20 +0000"},{"title":"Designing an Always-On Agent","summary":"Argues today's agents do nothing until messaged and explores what a continuously running, environment-aware agent needs.","source":"hackernews_ai","url":"https://rubriclabs.com/blog/always-on-agents","published":"Sun, 08 Nov 2026 17:51:44 +0000"}]},{"name":"Agent memory and context become shared infrastructure","slug":"agent-memory-context-infrastructure","summary":"Memory is fragmented per tool, and new projects try a shared format or an inspectable filesystem-style store.","articles":[{"title":"Agent memory needs its OpenTelemetry moment","summary":"Each coding tool stores memory in its own shape and none can read the others; the post surveys existing formats and sketches a shared spec.","source":"hackernews_ai","url":"https://www.lorekit.io/blog/open-agent-memory-format","published":"Sun, 08 Nov 2026 08:40:23 +0000"},{"title":"OpenViking: a context database for agents","summary":"Open-source store that exposes knowledge, memory and skills as a viking:// virtual filesystem agents navigate with ls, tree, read and grep.","source":"hackernews_ai","url":"https://github.com/volcengine/OpenViking","published":"Sun, 08 Nov 2026 04:44:14 +0000"},{"title":"RepoMagpie: your coding agent remembers repos you've tried","summary":"Keeps your verdicts on libraries, CLIs and MCP servers so the agent checks them before installing a tool you already ruled out.","source":"hackernews_ai","url":"https://github.com/Varshavia/RepoMagpie","published":"Sun, 08 Nov 2026 10:13:24 +0000"},{"title":"Google's Elephant-Goldfish Model plugin","summary":"Splits work in two: Elephant keeps project context and drafts the spec; Goldfish reviewers read it without the prior conversation to expose blind spots.","source":"hackernews_ai","url":"https://github.com/google/elephant-goldfish-model","published":"Sun, 08 Nov 2026 03:50:34 +0000"},{"title":"aiport: a local LLM gateway","summary":"Go gateway that routes Claude Code, Codex, Ollama and cloud models through one endpoint and shows which app called what.","source":"hackernews_ai","url":"https://github.com/harelix/aiport","published":"Sun, 08 Nov 2026 11:44:16 +0000"}]},{"name":"Security and reliability of agents under test","slug":"agent-security-reliability","summary":"Offensive agent use is cheap and measurable, while new work targets model backdoors and agent-loop failure rates.","articles":[{"title":"Check Point: AI agents stole 190k gaming credentials for under $400","summary":"Analysis of ARTEX, an open-source LLM-driven pentesting platform, and how cybercrime campaigns used it at low model cost.","source":"hackernews_ai","url":"https://research.checkpoint.com/2026/hacking-on-autopilot-inside-artex-cybercrime-campaigns/","published":"Sun, 08 Nov 2026 01:24:19 +0000"},{"title":"What Can AI Hack? Inside Frontier Labs' Tests","summary":"Explains how to read system-card cyber evals, including OpenAI rating GPT-6 Astra its first Critical-level cyber model in September 2026.","source":"hackernews_ai","url":"https://dorcohen.dev/posts/what-can-ai-actually-hack/","published":"Sun, 08 Nov 2026 09:42:03 +0000"},{"title":"Uncensored LLMs in agent loops: 31 models, 5 tasks","summary":"Most models still fail about half the time on multi-step tasks; models built on weaker bases loop almost twice as often.","source":"hackernews_ai","url":"https://openuncensored.com/blog/uncensored-models-in-agent-loops","published":"Sun, 08 Nov 2026 01:24:17 +0000"},{"title":"NC State: new techniques to prevent AI backdoor attacks","summary":"Three techniques to block embedded triggers, plus a proposal to reuse backdoor methods as a security layer for users.","source":"hackernews_ai","url":"https://news.ncsu.edu/2026/11/new-techniques-to-prevent-backdoor-attacks/","published":"Sun, 08 Nov 2026 09:32:43 +0000"}]},{"name":"Inference and fine-tuning off the CUDA path","slug":"inference-off-cuda-path","summary":"Two items show serving and tuning stacks moving beyond Nvidia and Python.","articles":[{"title":"VindexLLM: GGUF inference in Delphi on Vulkan","summary":"Runs full transformer inference on the GPU through Vulkan compute shaders, with no Python or CUDA, needing only vulkan-1.dll.","source":"hackernews_ai","url":"https://github.com/tinyBigGAMES/VindexLLM","published":"Sun, 08 Nov 2026 05:25:50 +0000"},{"title":"Huawei Ascend adapts Kimi K3.1 Pro for LoRA fine-tuning","summary":"Fine-tuning runs on a 128-card Atlas 950 SuperPoD, per Pandaily.","source":"search_cn_open_weight_labs","publisher_name":"Pandaily","publisher_domain":"pandaily.com","url":"https://news.google.com/rss/articles/CBMijAFBVV95cUxOTS1oQ3R3c2pRRUNhc3pTWmtRWkJueEtPVVZEa3RrUENjY0ZOTGE2SmVOZEN1YWtlWWlwUWJuR19fdTc1LUxMRjZRcFYwQ0MwdjdHSW5fTDdGU0RzYW11R0lVREtfVklaa1JZdEhnTkk2bEZPRXNoLThtdzRDanFHdkdrZEZqNWYwNTdXMA?oc=5","published":"Sun, 08 Nov 2026 01:50:51 GMT"}]}],"video":{"youtube_id":"xkKphBtXB_A","title":"LLM Digest Daily Explainer: Nov 8, 2026","duration_s":117.8}}